The one business all four AIs recommended is not that business any more.
For Edition 2 we re-measured an industry from Edition 1 after almost three months. The result: agreement fell from 1 to 0. Not because a restaurant did anything wrong, but because one AI changed its mind.
🇩🇪 Diesen Index auf Deutsch lesenThe KIGEO AI Visibility Index measures a single number: how often all four major AIs recommend the same Vienna business. Edition 1 set the baseline. Edition 2 shows for the first time what happens to that number over time. And that is the real story.
In Edition 1, exactly one of 31 Vienna restaurants was named by all four AIs. In the re-measurement almost three months later it is zero out of 24. Across five industries and roughly 134 recommended businesses, agreement stays close to zero.
The one number: agreement
The core question of the index is simple: how often do all four AIs name the same business? The rarer that is, the more your visibility depends on which AI your customer happens to ask.
| Industry | Collected | Named by AI | Named by all four |
|---|---|---|---|
| Restaurants | June 2026 | 31 | 1 |
| Restaurants, re-measured | Sept 2026 | 24 | 0 |
| Tax advisors | June 2026 | 30+ | 0 |
| Vets | June 2026 | 16 | 0 |
| Real estate agents (6 districts) | July 2026 | ~34 | in 3 of 6 districts |
| Dentists, general question | Aug 2026 | 11 | 0 |
| Dentists, buying question | Aug 2026 | 13 | 0 |
Bold marks what is new in Edition 2. For real estate agents we measured per district, for every other industry city-wide.
What changed since Edition 1
On 4 September 2026 we repeated the restaurant question from Edition 1 word for word, in the same four engines:
(“What are the best restaurants in Vienna? Please give me specific recommendations.”)
In Edition 1, Steirereck was the only business in all of Vienna named by all four AIs. In the re-measurement it stands at three out of four. Google, ChatGPT and Gemini still name it. Perplexity does not.
The reason has nothing to do with the restaurant. Perplexity simply read the question differently. In June it answered with fine dining and classics: Steirereck, Amador, Silvio Nickol, Figlmüller. In September it names everyday places, including a noodle shop and a bowl bar. Not one fine dining house left.
An AI recommendation is not something you own. It can disappear without you changing a thing.
How far apart the four sit shows in the pairwise comparison of that same round of answers:
| Pair | restaurants named in common |
|---|---|
| Google AI Mode and ChatGPT | 5 |
| ChatGPT and Gemini | 4 |
| Google AI Mode and Gemini | 2 |
| Perplexity and Google AI Mode | 1 |
| Perplexity and ChatGPT | 0 |
| Perplexity and Gemini | 0 |
The sources shifted too. ChatGPT relied on Google Maps and Reddit for this question in June, and on Michelin, Gault Millau, Falstaff and vienna.info in September. Perplexity cited restaurant guides in June and now almost only the venues themselves plus map data. Whoever wants to win an engine is aiming at a moving target.
New in Edition 2: the buying question
The dentist industry adds a new dimension to the index. Until now we always asked generally (“a good dentist in Vienna”). This time we also asked, on the same day, the way someone asks who is about to spend money (“I need a dental implant”).
Of 17 recommendations from the general question, only 2 survived the switch to the buying question. For ChatGPT and Gemini it was none at all. Read the study →
Why do they disagree so much?
Because each AI reads different sources, and because those sources change. Across all our measurements so far:
- ChatGPT and Google usually read Google Maps and reviews, but switch to specialist publications and guides for editorial topics.
- Perplexity is the most volatile case. Depending on the industry it cites business directories, the venues' own websites, or map data. For dentists it was 10 out of 10 practice websites; for restaurants in September, almost only map data.
- Gemini shows Google Maps pins with star ratings and almost never names a source.
The logic stays the same: no source, no mention. If you are missing from an engine's sources, that engine will not recommend you, however good the business is. This matches larger studies: ChatGPT names only around 1 in 100 local businesses at all1, and the engines draw their recommendations from almost entirely different sources2.
The five industries in detail
Restaurants. 31 venues in June, one named by all four. In the September re-measurement, 24 venues and none left. Read the study →
Dentists. 11 practices on the general question, 13 on the implant question, none in all four. Only 2 of 17 recommendations survive the switch to the buying question. Read the study →
Tax advisors. More than 30 firms, not one in all four. The AIs split into two camps depending on whether they read reviews or directories. Read the study →
Vets. 16 practices, none in all four, and roughly one in four recommendations carried an error. This one was about accuracy, not only visibility. Read the study →
Real estate agents. Six districts, six completely different lists. Where a firm won everywhere, it always had both: plenty of fresh reviews and clean directory entries. Read the study →
What this means for businesses
- Visibility differs per AI. A strong Google profile gets you into ChatGPT, Google and Gemini. For Perplexity you need directories or a usable website of your own, depending on the industry. One channel is not enough.
- Visibility is not permanent. The Steirereck case shows a mention can vanish without the business changing anything. Measuring once tells you nothing about today.
- The general question and the buying question are two separate jobs. One is won by the profile, the other by explainer content about the specific service.
That is exactly what Generative Engine Optimization (GEO) is for: making sure you get named in every relevant AI, in the right area, on the right question and with correct details.
Does AI recommend your business?
We check for free whether ChatGPT, Perplexity, Gemini and Google name you in Vienna, and show who gets recommended instead.
get a free AI visibility check →Frequently asked questions
What is the KIGEO AI Visibility Index?
The index measures how often the four major AI search engines ChatGPT, Perplexity, Gemini and Google recommend the same business. Edition 2 covers five Vienna industries with roughly 134 recommended businesses and re-measures one industry after three months for the first time. It is published quarterly.
What is the result of Edition 2?
In the restaurant re-measurement, not one of the 24 addresses was named by all four AIs. In Edition 1 there was still one, Steirereck. It lost that status because Perplexity now answers the question differently, not because anything changed at the restaurant.
Does an AI recommendation stay once you are in?
No. Our re-measurement after almost three months shows a business can lose its mention without changing anything itself. The engines change which sources they read and how they interpret a question.
Why do the AI search engines disagree so much?
Because each reads different sources. ChatGPT and Google mostly rely on Google Maps and reviews, sometimes on specialist guides; Perplexity on directories, own websites or map data depending on the industry; Gemini on Maps pins with star ratings. No source, no mention.
How do I get into all four AIs?
By appearing cleanly in each engine's sources: plenty of fresh reviews for ChatGPT, Google and Gemini, plus consistent directory entries and a usable website of your own for Perplexity. And by serving not only the general question but the concrete buying question too. That is exactly what Generative Engine Optimization (GEO) does.
How was the index measured?
The same word-for-word question per industry from a Vienna browser on free access, one fresh chat per engine, recording every business named. Period June to September 2026. These are samples, not full surveys, and AI answers fluctuate.
When is the next edition?
Edition 3 is planned for the first quarter of 2027. We will add further industries and re-measure an existing one again, so the development over time stays readable.
Methodology & sources
Edition 2 draws on the published individual studies, whose raw data is available there as CSV, plus the newly collected restaurant re-measurement of 4 September 2026. Studies used for context:
- SOCi, Local Visibility Index 2026, context: across hundreds of thousands of locations, ChatGPT names only around 1.2 percent of local businesses in a recommendation (US dataset, indicative).
- NeuRank, AI visibility study 2026, context: ChatGPT bases roughly 38 percent of its local recommendations on Google Maps, while Perplexity cites almost exclusively German-language directories (German market).
- Own data collection: KIGEO, restaurant re-measurement Vienna, 4 September 2026, four engines, word-for-word identical question.